CODEWITHSUNDEEP
pythondataframepandas
william
william

Reputation: 31

Pandas Dataframe label selection

Can anybody tell me why df['2005-5-31'] trigger the KeyError exception?

rng = pd.date_range('2005', '2012', freq='M')
df = pd.DataFrame(randn(len(rng), 3), rng, ['X', 'Y', 'Z'])
df.head()

# works
df.ix['2005-5-31']
df['2005-5-31':'2005-5-31']

# Gives KeyError: u'no item named 2005-5-31'
df['2005-5-31']

Follow code using df['2000-01-01'] works.

#multiple rows on a single date
rng = pd.date_range('2000-01-01', '2000-01-3', freq='8H')
df = pd.DataFrame(randn(len(rng), 3), rng, ['X', 'Y', 'Z'])

# works
df['2000-01-01']

                           X    Y   Z
2000-01-01 00:00:00 -0.227981    1.927932   -0.518947
2000-01-01 08:00:00  0.486063   -1.255186    0.375075
2000-01-01 16:00:00 -2.313950    0.654384    1.111493

Upvotes: 1

Views: 266

Answers (2)

waitingkuo
waitingkuo

Reputation: 93924

df['2005-5-31'] is to select by column, but you don't have a column name 2005-5-31.

df['X'] works since you have a columns name x

Edit

The reason df['2005-5-31':'2005-5-31'] is to select by index but not column is because it doesn't make sense to make a slicing selection on columns.

For more information, take a look at here

Upvotes: 1

Ryan Saxe
Ryan Saxe

Reputation: 17869

df['2005-5-31'] returns the column that is named: 2005-5-31. Your columns are named X,Y,Z. And because you don't have the date column, it is giving you an error!

Now the .ix[] method works because it takes up to two inputs, the first input being the row index. You have a row with the index '2005-5-31' so it works!

Upvotes: 1

Related Questions